🌿 An algo to get rid of pot convictions and the EU is not here to play games with dangerous AI

🎭 Deepfakes for fun!

The Good

In 2016, California decriminalized recreational marijuana use and made the law retroactive. Finding and fixing the hundreds of thousands of convictions takes a lot of time, however, deferring justice for many who have been waiting for years now.

But an algo could make this all move much faster.

Which Los Angeles is proving with its partnership with Code for America, which used its Clear My Record tech to find 66,000 (!) convictions dating back to 1961 to dismiss.

Normally, when we here about algos in criminal justice settings it is Very Bad News. In fact, this might be the first positive case I’ve read about and as far as I can tell, this isn’t even machine learning. It would be interesting to see if an even more advanced algo could do more good in this situation or ones like it.

The Better

isn’t this a nice change for once?

The European Commission announced they want a “human-centric” way forward with AI, specifically calling for a different way forward than China and America’s cutthroat mentality.

The Commission wants strict rules around AI applications that have a high risk of harm if things go wrong—like in healthcare, transportation, and criminal justice. For these uses, black boxes, or tools that are inscrutable even to the people who made them, would be banned. When it comes to face rec tech, the group wants a debate within the EU about how the citizens want to go forward.

This press release has so many ideas and info it makes me think that the EU is the Elizabeth Warren of governments when it comes to AI: They have a plan for that.

I think instead of putting the EU behind in the “AI race,” this approach could put them ahead. Some of these tech use cases have to get it right the first time (like truly autonomous vehicles) or people won’t trust/buy/use it. Slowly down the process and making sure AI works for people is a way to do that. It’s also setting out ground rules that, if in place early enough, could force other companies to abide by if they want access to the EU market.

More News

I am here for this Office Space meets Matrix deepfake.

Another deepfake! Back to the Future with Tom Holland and Robert Downey Jr. switched in.

One last deepfake: an Indian politician made to look like he was speaking in different languages to reach more voters.

The U.S. chief tech officer, who is 33 with a Greek history degree, thinks the EU’s AI plan will let “China’s digital authoritarianism” get ahead. I am only slightly dunking on him since I am 34 with an American history degree and also have ~opinions~ about AI. Except mine doesn’t shape U.S. policy.

AI is counting seals and belugas from space to help save the species despite the melting Arctic.

More details about a crash that involved Tesla’s AutoPilot (which, reminder, is not actually supposed to be used as an autopilot) and how it was pretty traumatizing for the truck driver that the Tesla hit.

Bonus Musk content: he thinks AI should be regulated, even Tesla.

A little nerdy but could have a wide-ranging impact: Google got rid of gendered tags for its computer vision API that a lot of different developers use.

Developers released an AI-powered tool to block dick pics in Twitter DMs. Sure, we are going to have to jeopardize our privacy but with 3 out of 4 millennial ladies getting these schnitzels shots, what do we do?

***

Until we get a deepfake that lets me put Lee Pace into any movie I want,

Jackie